贪心非对称深度有监督哈希图像检索方法  

Greedy-asymmetric deep supervised hashing for image retrieval

作　　者：赵昕昕 李阳 苗壮 王家宝 张睿 Zhao Xinxin;Li Yang;Miao Zhuang;Wang Jiabao;Zhang Rui(College of Command&Control Engineering,Army Engineering University of PLA,Nanjing 210007,China)

机构地区：[1]陆军工程大学指挥控制工程学院,南京210007

出　　处：《计算机应用研究》2022年第10期3156-3160,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61806220);江苏省自然科学基金资助项目(BK20200581)。

摘　　要：近年来,深度有监督哈希检索方法已成功应用于众多图像检索系统中。但现有方法仍然存在一些不足:一是大部分深度哈希学习方法都采用对称策略来训练网络,但该策略训练通常比较耗时,难以用于大规模哈希学习过程;二是哈希学习过程中存在离散优化问题,现有方法将该问题进行松弛,但难以保证得到最优解。为解决上述问题,提出了一种贪心非对称深度有监督哈希图像检索方法,该方法将贪心算法和非对称策略的优势充分结合,进一步提高了哈希检索性能。在两个常用数据集上与17种先进方法进行比较。在CIFAR-10数据集上48 bit条件下,与性能最好的方法相比,mAP提高1.3%;在NUS-WIDE数据集上所有比特下,mAP平均提高2.3%。在两个数据集上的实验结果表明,该方法可以进一步提高哈希检索性能。In recent years,the deep supervised hash retrieval method has been successfully applied to many image retrieval systems.However,the existing methods still have some shortcomings.Firstly,most of the deep hash learning methods used symmetric strategies to train the network,but the training of this strategy was usually time-consuming and difficult to be used in the large-scale hash learning process.Secondly,there was a discrete optimization problem in the hash learning process.Exis-ting methods relaxed this problem and it was difficult to guarantee the optimal solution.In order to solve the above problems,this paper proposed a greedy-asymmetric deep supervised hashing method for image retrieval,which fully combined the advantages of the greedy algorithm and asymmetric strategy to further improve the hash retrieval performance.This article compared 17 state-of-the-art methods on two commonly used datasets.Compared with the state-of-the-art methods,this proposed method increased the mAP in 48 bit setting by 1.3%on CIFAR-10 dataset.And on NUS-WIDE dataset,it increased the mAP in all-bits setting by increased 2.3%on average.The experimental results show that this proposed method can further improve the performance of hash retrieval.

关 键 词：非对称策略 贪心算法 有监督哈希 图像检索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]